ADC Input readings propagate between channels on RM46

Other Parts Discussed in Thread: HALCOGEN, TINA-TI

Hey everyone,

I am using the ADC modules on a RM46 to sample a few analog inputs. These are custom PCBs and I have attached the important parts of the electrical schematic. The issue I am seeing is a small cross over between ADC channels in the same group that is proportional to and input channel. For example the figure below is a capture of AIN[2], AIN[3] and AREFHI pins to the RM46L852-PGET chip. When I increase the voltage going into AIN[2] you can see an increase in the noise on AIN[3]. This translates into readings crossing the channel. With AIN[2] at ~0 AIN[3] is ~0 but when I increase AIN[2] to ~3600 AIN[3] increases to ~316. BUT this does not work in reverse, increasing the input into AIN[3] does not effect AIN[2] like I would expect:

After some debugging I found that the noise corresponds to my ADC samples:

There are breaks visible when the CPU is not doing ADC conversions, like writing to CAN or SCI. In my software I am running this:

void main()
{
    	adcInit();

    	while(1)
    	{
    	    	adcData_t d[ADC_MAX_CHANNEL_COUNT];

	    	adcStartConversion( adcREG1, adcGROUP1 );
	    	// wait while ADC is converting
	    	while( !adcIsConversionComplete( adcREG1, adcGROUP1 ) );
	    	// Get the data
	    	adcGetData( adcREG1, adcGROUP1, &d[0]);

	    	brakePressureMonitor.Raw = d[0].value;
	    	tsVehicleVoltageMonitor.Raw = d[1].value;
	    	driveThrottleOneMonitor.Raw = d[2].value;
	    	driveThrottleTwoMonitor.Raw = d[3].value;

    	    	// Other stuff afterwords eg. CAN transmissions, SCI transmissions
    	}
}

Schematics:

Any ideas what could be causing this behavior? If you would like any more information let me know!

Thanks,

Nathan